Search for: All records

The formation of boundaries separating developmental fields with distinct gene expression and cell-fate trajectories is a universal feature of non-colonial multicellular organisms. Developmental boundaries arise reiteratively during ontogeny and are characterized by stiff, slowly dividing cells that demarcate adjacent and divergent morphogenetic domains; the genetic mechanisms of cell-fate acquisition within these boundaries are incompletely understood. Grass leaves are initiated at a developmental boundary in the periphery of the shoot apical meristem (SAM), an organogenic pool of plant stem cells that generates all lateral organs in the plant shoot. During later primordial growth, maize leaves form a de novo developmental boundary that ultimately separates the distal, photosynthetic leaf blade from the proximal, clasping leaf sheath. Morphogenesis at this blade/sheath boundary in maize leaves generates an epidermal outgrowth called the ligule and two tissue-wedges forming the auricle, a hinge-like structure with major effects on leaf angle, light capture, and yield. Here, we use cell-lineage mapping, morphometric measures of cell division and expansion, cell-specific multidimensional transcriptomic analyses, and topological landscape modeling to investigate the mechanisms of cell-fate acquisition at the ligule/auricle morphogenetic boundary in the maize leaf. The data suggest a model where auricle initial cells are recruited from blade founder cells at this boundary, via repression of blade identity during early stages in auricle ontogeny. Thereafter, auricle primordial cells assume a developmental genetic trajectory that is distinct from the blade, sheath, and ligule, thereby acquiring a unique auricle cell-fate in the maize leaf. 

Abstract The Scintillating Bubble Chamber (SBC) Collaboration is designing a new generation of low background, noble liquid bubble chamber experiments with sub-keV nuclear recoil threshold. These experiments combine the electronic recoil blindness of a bubble chamber with the energy resolution of noble liquid scintillation, and maintain electron recoil discrimination at higher degrees of superheat (lower nuclear recoil thresholds) than Freon-based bubble chambers. A 10-kg liquid argon bubble chamber has the potential to set world leading limits on the dark matter nucleon cross-section for 𝒪(GeV/c²) masses, and to perform a high statistics coherent elastic neutrino nuclear scattering measurement with reactor neutrinos. This work presents a detailed calibration plan to measure the detector response of these experiments, combining photoneutron scattering with two new techniques to induce sub-keV nuclear recoils: nuclear Thomson scattering and thermal neutron capture. 

Recent approaches have shown promises distilling diffusion models into efficient one-step generators. Among them, Distribution Matching Distillation (DMD) produces one-step generators that match their teacher in distribution, without enforcing a one-to-one correspondence with the sampling trajectories of their teachers. However, to ensure stable training, DMD requires an additional regression loss computed using a large set of noise-image pairs generated by the teacher with many steps of a deterministic sampler. This is costly for large-scale text-to-image synthesis and limits the student's quality, tying it too closely to the teacher's original sampling paths. We introduce DMD2, a set of techniques that lift this limitation and improve DMD training. First, we eliminate the regression loss and the need for expensive dataset construction. We show that the resulting instability is due to the fake critic not estimating the distribution of generated samples accurately and propose a two time-scale update rule as a remedy. Second, we integrate a GAN loss into the distillation procedure, discriminating between generated samples and real images. This lets us train the student model on real data, mitigating the imperfect real score estimation from the teacher model, and enhancing quality. Lastly, we modify the training procedure to enable multi-step sampling. We identify and address the training-inference input mismatch problem in this setting, by simulating inference-time generator samples during training time. Taken together, our improvements set new benchmarks in one-step image generation, with FID scores of 1.28 on ImageNet-64x64 and 8.35 on zero-shot COCO 2014, surpassing the original teacher despite a 500X reduction in inference cost. Further, we show our approach can generate megapixel images by distilling SDXL, demonstrating exceptional visual quality among few-step methods. 

We propose a novel model-based reinforcement learning algorithm—Dynamics Learning and predictive control with Parameterized Actions (DLPA)—for Parameterized Action Markov Decision Processes (PAMDPs). The agent learns a parameterized-action-conditioned dynamics model and plans with a modified Model Predictive Path Integral control. We theoretically quantify the difference between the generated trajectory and the optimal trajectory during planning in terms of the value they achieved through the lens of Lipschitz Continuity. Our empirical results on several standard benchmarks show that our algorithm achieves superior sample efficiency and asymptotic performance than state-of-the-art PAMDP methods.